多方法融合来解决人脸检测中的光照补偿  

Illumination Compensation Based on Multi-method Integration in Face Detection

作　　者：刘宏[1] 李锦涛[1] 苗军[1] 

机构地区：[1]中国科学院计算技术研究所数字化室 ICT-YCNC面像识别联合实验室,北京100080

出　　处：《系统仿真学报》2001年第S2期486-489,510,共5页Journal of System Simulation

基　　金：国家自然科学基金(编号:69789301);成都银辰网讯的赞助。

摘　　要：人脸检测有广泛的应用前景,但目前的许多检测算法对光照有一定的要求,过亮、过暗或偏光现象的存在使检测率急剧下降。本文提出一种多方法融合的思想来解决光照补偿问题,并给出了基于空域和频域的几种具体的解决方案。实验表明,这种多方法融合的思想能有效地解决人脸检测中的光照补偿问题,使人脸检测率得到明显的提高。Face detection has wide applications. But some of the detection algorithms are sensitive to illumination variation. Detection rates usually drop quickly if the illumination is too bright, too dark, or non-uniform illumination. In this paper, an idea of multi-method integration is proposed to solve the illumination compensation problem in face detection. Some detailed methods based on frequency-domain and spatial-domain are also introduced in this paper. The experiment data prove that the idea of multi-method integration can well overcome the illumination compensation problem and the rate of face detection is obviously increased.

关 键 词：人脸检测 光照补偿 非线性变换 同态滤波 熵 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]